chemistry & biology, vol. 11, 1179–1194, september, 2004, 2004 … · 2016-12-01 ·...

Chemistry & Biology, Vol. 11, 1179–1194, September, 2004, 2004 Elsevier Ltd. All rights reserved. DOI 10.1016/j .chembiol .2004.05.024

An Aldol Switch Discovered in StilbeneSynthases Mediates Cyclization Specificityof Type III Polyketide Synthases

differ in their choice of CoA-tethered starter molecule,in the number of polyketide elongation steps catalyzed,and in the mechanism for intramolecular cyclization oflinear polyketide intermediates, thereby producing a va-riety of biologically and medicinally important natural

Michael B. Austin,1,2 Marianne E. Bowman,1

Jean-Luc Ferrer,3 Joachim Schroder,4

and Joseph P. Noel1,2,*1Structural Biology LaboratoryThe Salk Institute for Biological Studies10010 North Torrey Pines Road product classes [2]. While the enzymatic features that

mediate starter molecule selection and polyketide chainLa Jolla, California 920372 Department of Chemistry and Biochemistry extension are largely understood, type III PKS control

of cyclization specificity remains a mystery.University of California, San DiegoLa Jolla, California 92037 The classic illustration of this latter point contrasts

the first two examples of type III PKS to be discovered:3 IBS J.-P. Ebel41 rue Jules Horowitz the chalcone synthase (CHS) and stilbene synthase (STS)

enzyme families [3]. STS has independently evolved in38027 Grenoble Cedex 1France a few diverse plants (such as grapevine, pine, and pea-

nut) via the duplication and divergence of chs genes,4 Institut fur Biologie IIBiochemie der Pflanzen the latter of which are ubiquitous in higher plants [4].

Despite their limited occurrence in nature, the di- andUniversitat FreiburgSchanzlestrasse 1 trihydroxylated stilbene scaffolds produced by STS

have recently received much attention due to their nu-D-79104 FreiburgGermany merous biological activities. Expression of sts genes

confers significant resistance against fungal infectionto both natural and heterologous host plants [5, 6]. Res-veratrol (trihydroxy-stilbene) is also believed to be aSummarymajor contributor to the health benefits associated withthe moderate consumption of red wine (known as “theStilbene synthase (STS) and chalcone synthase (CHS)French paradox”) [7]. Indeed, animal cell culture studieseach catalyze the formation of a tetraketide intermedi-have linked an impressive number of beneficial medici-ate from a CoA-tethered phenylpropanoid starter andnal effects to resveratrol and other stilbenes, includingthree molecules of malonyl-CoA, but use different cy-copper chelation, antioxidant scavenging of free radi-clization mechanisms to produce distinct chemicalcals, inhibition of both platelet aggregation and lipidscaffolds for a variety of plant natural products. Hereperoxidation, antiinflammatory activity, vasodilation, andwe present the first STS crystal structure and identify,anticancer activities [8, 9]. Most recently, resveratrolby mutagenic conversion of alfalfa CHS into a func-was found to delay apoptosis through a 13-fold stimula-tional stilbene synthase, the structural basis for thetion of the sirtuin-family deacetylase activity that nega-evolution of STS cyclization specificity in type IIItively regulates the p53 tumor suppressor, resulting inpolyketide synthase (PKS) enzymes. Additional muta-the same significant extension of organismal lifespangenesis and enzymatic characterization confirms thatpreviously observed in connection with severe restric-electronic effects rather than steric factors balancetion of caloric intake [10].competing cyclization specificities in CHS and STS.

CHS and STS enzymes share 75%–90% amino acidFinally, we discuss the problematic in vitro reconstitu-sequence identity over their �400 residues, and both en-tion of plant stilbenecarboxylate pathways, using in-zyme families catalyze the same iterative condensationsights from existing biomimetic polyketide cyclizationof three acetyl units (derived from the decarboxylationstudies to generate a novel mechanistic hypothesis toof malonyl-CoA) to a CoA-tethered phenylpropanoidexplain stilbenecarboxylate biosynthesis.starter molecule (derived from phenylalanine), most typi-cally p-coumaroyl-CoA. However, STS enzymes cyclize

Introduction the resulting tetraketide intermediate product via an in-tramolecular C2→C7 aldol condensation, rather than the

Type III polyketide synthases (PKSs) are structurally intramolecular C6→C1 Claisen condensation utilized bysimple, homodimeric iterative PKSs (Figure 1A) that uti- CHSs (Figure 1D) [11].lize a conserved Cys-His-Asn catalytic triad in an internal Previously, the alfalfa CHS crystal structure impliedactive site (Figure 1B) to catalyze the iterative condensa- that proper orientation of CHS’s enzyme-bound tetrake-tion of acetyl units (derived from malonyl-CoA) to a CoA- tide intermediate by residues lining the internal activelinked starter molecule (Figure 1C). This chain extension site cavity is sufficient to promote the CHS intramolecu-is usually followed by cyclization of the linear polyketide lar Claisen cyclization reaction [12]. This C6→C1 cycliza-intermediate in the same active site cavity [1, 2]. The tion also conveniently results in cleavage of the C1 thio-more than a dozen functionally distinct plant and bacte- ester linkage to the CHS enzyme, thus offloading therial type III PKS enzyme families characterized to date chalcone product. Conversely, the STS C2→C7 reaction

requires a thioesterase-like hydrolysis step to cleavethe C1 thioester linkage to the STS enzyme, as well as*Correspondence: [email protected]

Chemistry & Biology1180

Figure 1. Type III Polyketide Synthase Enzymes and Tetraketide Cyclization

(A) Type III PKS homodimeric architecture (blue and gold) and CoA binding based upon alfalfa chalcone synthase (CHS) crystal structures[12]. Bound CoA (stick model) highlights the entrance to the CoA binding tunnel for each monomer’s internal active site cavity (red box: see[B]). Catalytic cysteine positions are noted (blue asterisks).(B) CHS active site cavity occupied by chalcone-derived naringenin (cyan), shown from the perspective of the CoA binding tunnel. Shown arethe condensing machinery (Cys-His-Asn catalytic triad and Phe215) in rose, additional residues lining the active site cavity (blue), and theonly active site cavity residue contributed by the dyad-related monomer (gold).(C) Type III PKS condensing mechanism (starter loading and polyketide extension). CoA-linked starter moieties (green) are covalently loadedonto the catalytic cysteine. The decarboxylation of malonyl-CoA produces an activated acetyl unit (pink) that undergoes Claisen condensationwith the cysteine-bound starter moiety. The resulting diketide intermediate is transferred from CoA to the catalytic cysteine for additionaltwo-carbon extension reactions.(D) Identical tetraketide intermediates produced by CHS and STS (stilbene synthase) reactions, from p-coumaroyl-CoA (R � OH) or cinnamoyl-CoA (R � H) starter molecules. This tetraketide intermediate initially forms as a CoA-thioester, rather than the enzyme-linked thioester depictedhere. Alternative intramolecular cyclization patterns lead to different natural product scaffolds. The blue arrow depicts CHS’s C6→C1 Claisencondensation leading to hydroxylated chalcones, the red arrow shows the C2→C7 aldol condensation leading to the STS-synthesizedhydroxylated stilbenes discussed here (resveratrol [R � OH] and pinosylvin [R � H]), and the green arrow illustrates the C5-oxygen→C1lactonization associated with p-coumaroyl tetraacetic acid synthase (CTAS, see text), as well as product derailment in both CHS and STS.Finally, the still unresolved reaction leading to stilbene acids is depicted.

Stilbene Synthase Crystal Structure and Mechanism1181

Table 1. Crystallographic Data and Refinement Statistics

STS 18xCHS 18xCHS � Resveratrol

Space group P2(1) P2(1) P1Unit cell dimensions (A,�) a � 57.2, b � 361.3, a � 71.6, b � 59.8, a � 64.3, b � 71.7, c � 85.7,

c � 57.3, � � 98.4 c � 82.5, � � 108.2 � � 111.4, � � 91.6, � � 90.1Wavelength (A) 0.933 0.931 0.773Resolution (A) 2.1 1.9 2.0Total reflections 202,864 186,828 189,393Unique reflections 103,258 52,432 93,036Completenessa (%) 74.1 (34.1) 98.7 (99.9) 97.5 (96.8)I/�a 4.2 (1.6) 13.6 (3.8) 14.1 (2.2)Rsym

a,b 18.2 (30.7) 4.4 (17.0) 5.8 (30.4)Rcryst

c/Rfreed (%) 22.5/28.9 19.3/23.0 20.3/26.4

Protein atoms 17,892 5,881 11,896Ligand atoms — — 69Water molecules 1871 716 1190Rmsd bond lengths (A) 0.008 0.006 0.010Rmsd bond angles (�) 1.4 1.3 1.5Average B factor: protein (A2) 34.9 20.4 31.1Average B factor: solvent (A2) 69.0 67.0 51.0

a Number in parenthesis is for the highest resolution shell.b Rsym � �|Ih Ih�|/�Ih, where Ih� is the average intensity over symmetry equivalent reflections.c R factor � �|Fobs Fcalc|/�Fobs, where summation is over the data used for refinement.d Rfree factor is the same definition as for R factor, but includes only 5% of data excluded from refinement.

an additional decarboxylative elimination of the resulting into a functional stilbene synthase confirms our identi-fication of the “aldol switch” structural changes respon-C1 carboxylate. However, sequence comparison of the

STS subfamilies both to each other and to CHS reveals sible for aldol cyclization specificity. Further subtlemutations designed to perturb only the proposed thioes-no apparent STS consensus sequence [4], and homol-

ogy modeling carried out in our lab on STS enzymes terase step support our conclusions. Finally, we offera novel mechanistic proposal for the biosynthesis ofsubsequent to determination of the alfalfa CHS2 crystal

structure predicted no significant topological or chemi- stilbenecarboxylic acids that draws circumstantial sup-port from a number of biomimetic polyketide cyclizationcal differences in the STS active site cavities relative to

CHS. studies.In the absence of any convincing STS sequence-

related or structural evidence, it has been presumed that Results and Discussionsome steric reshaping of the active site cavity, relativeto CHS, directs the divergent C2→C7 aldol cyclization 2.1 A Crystal Structure of Pinosylvin-Forming

Stilbene Synthase from P. sylvestrisspecificity of STS. While this steric modulation hypothe-sis provides a reasonable explanation for how STS could The unusual length (361.3 A) of one of the crystallo-

graphic axes of the pine STS unit cell translates intoachieve an alternative productively folded conformationof the linear tetraketide intermediate it shares with CHS reflections that are closely spaced and overlapped. De-

convolution of these overlaps represented a significant[12, 13], this model fails to account for the additional STSthioesterase activity and polyketide C1 decarboxylation data processing obstacle, similar to problems encoun-

tered in crystallographic analyses of virus structures,steps required for biosynthesis of the stilbene scaffold.Notably, no structural feature of STS related to catalysis that here impacts both the completeness and quality

of even our best pine STS data set (see Experimentalof the thioesterase activity has yet been identified oreven proposed. The timing and mechanistic relevance Procedures and Table 1). However, despite poorer sta-

tistics relative to most structures of similar resolution,of the C1 decarboxylation step also remains unresolved,although this process was recently determined to occur even the initial STS 2Fo Fc and Fo Fc electron density

difference maps were highly informative, clearly show-prior to aromatization of the stilbene product’s newlyformed dihydroxyphenyl ring moiety [14]. ing the structural differences between STS and the initial

CHS-derived STS homology model. Moreover, the struc-To illuminate the structural and mechanistic basis forthe intramolecular aldol cyclization specificity of STS tural and mechanistic conclusions gleaned from this first

STS crystal structure were independently verified by theenzymes, we here present our structural elucidation ofa pinosylvin-forming STS from Pinus sylvestris (Scots mutagenic conversion of polyketide cyclization specific-

ity achieved in STS structure-guided CHS mutants. Thepine) [15]. This first stilbene synthase crystal structuredisfavors the prevailing model that STS achieves aldol higher statistical quality of the subsequent apo and res-

veratrol-complexed crystal structures of one such mu-cyclization specificity through steric modulation of theactive site architecture and instead implicates an unan- tant enzyme (detailed below and in Table 1) further vali-

dates the value and accuracy of the pine STS crystalticipated chemical mechanism guided by the emer-gence of a cryptic thioesterase activity. Our subsequent structure.

As expected, Scots pine STS closely adheres to thestructure-guided mutagenic conversion of alfalfa CHS


Figure 2. STS Crystal Structures and Functionally Relevant Differences

(A) C-� trace of one monomer (gold) of the homodimeric pine STS 2.1 A crystal structure overlaid with the alfalfa CHS homodimer (gray andblue), shown here with naringenin bound in each CHS active site cavity. A red box highlights the area 2 loop that separates these juxtaposedactive site cavities. CHS residues 132 and 137, which contribute to dyad-related active site cavities, are also shown.(B) Close-up view of the conformationally divergent pine STS area 2 loop (with equivalent CHS numbering in blue), showing the final 2Fo

Fc electron density map (blue wirecage) contoured at 1 sigma.(C) Alignment table defining CHS-like, intermediate, or STS-like (respectively labeled 0, 1, or 2; see text for explanation) sequences in regionA (areas 1–3) and region B (area 4). The intermediate sequences represent our initial guesses of the functionally relevant differences betweenthe two wild-type sequences, and are also the mutations introduced in the first of two rounds of mutagenesis necessary to fully convert eachrelevant area in CHS to the STS-like sequence (see Experimental Procedures). CHS residues (unique or shared) are blue, while mutationscorresponding to the pine STS sequence are gold. Gold residues in the bottom row correspond to the 18 mutations comprising the 18xCHS(2222) alfalfa CHS mutant that functions as an authentic STS. Gold residues in the black boxes comprise the eight mutations in the STS-like8xCHS (1210) mutant.(D) C-� trace of the 1.9 A crystal structure of the resveratrol-producing 18xCHS mutant. Mutated positions corresponding to region A (areas1–3) and region B (area 4) are in gold. Side chains of positions 132 and 137 are shown for clarity.(E) Resveratrol-forming cyclization specificity of engineered stilbene synthases (8xCHS and 18xCHS), compared to the observed productspecificity of wild-type alfalfa CHS and pine STS enzymes, illustrated using a thin layer chromatography (TLC) analysis of identical reactionsusing p-coumaroyl-CoA as starter molecule. (CHS’s enzymatic chalcone product undergoes a facile Michael addition reaction in solution toform the flavanone naringenin.) Steady-state kinetic constants for each reaction (see text) are tabulated below. Units of kcat, KM, and kcat/KM

are min1, �M, and M1s1, respectively. Average values are shown (n � 3).(F) Five 8xCHS area 2 loop mutations shown on an overlay of wild-type alfalfa CHS (blue) and wild-type pine STS (gold) structures.(G) Three 8xCHS areas 1 and 3 compensatory mutations shown in overlay of wild-type CHS (blue) and wild-type STS (gold) structures.

same overall structural fold observed in the two pre- thiolase fold, is conserved among all thiolase and con-densing enzymes crystallized to date, exposing theirviously structurally characterized plant type III PKSs [12,

16]. The �� core of this fold, also known as the ancient evolutionary relationship [2]. Surprisingly, crys-


tallographic comparison of the pine STS active site alfalfa CHS to the corresponding residues in Scots pineSTS (see Figure 2C and Experimental Procedures). In-cavity to that of alfalfa CHS reveals only very minor

differences in topology, none of which seem capable of deed, the resulting 18xCHS mutant functionally resem-bled wild-type STS, as reflected by both its steady-statepromoting an alternate conformation of the two en-

zymes’ identical tetraketide intermediate product, as kinetic properties and product specificity, producingresveratrol rather than chalcone as the major producthad been theorized [12, 17]. Overall, the C-� trace of

the entire STS backbone superimposes almost perfectly of in vitro assays (Figure 2E).Next, a quasicombinatorial mutagenic PCR strategywith that of CHS (rmsd � 0.65 A), diverging significantly

in only two regions, defined as A and B (Figures 2A–2D). using alfalfa CHS as a template was devised to decon-volute the mechanistic contribution toward cyclizationIndeed, a comparison of these two conformationally

distinct STS loops to the equivalent loops in the CHS specificity of each of the four stretches of primary se-quence implicated by the 18xCHS mutant’s assay re-structure reveals a C-� rmsd of 2.0 A. Notably, these

backbone changes are conserved in each of the six sults. In each of these four areas, the alfalfa CHS se-quence either remained unchanged (designated by a 0),noncrystallographically related monomers (three physi-

ological dimers) that comprise the pine STS crystal’s or was partially (1) or completely (2) mutated to the Scotspine STS sequence (Figure 2C). These CHS mutantsunusually large asymmetric unit.

Region A spans a buried 6 residue loop, located at were labeled accordingly using a four-digit code to re-flect their composition in each area (for example, thethe dimer interface, that separates the two monomers’

identical active sites. Thr135 at one end of the loop 18xCHS mutant was labeled 2222, while wild-type CHSwould be labeled 0000). It was neither necessary norcontributes to the active site of its polypeptide chain

while Leu140 at the other end of the loop contributes efficient to construct all 81 possible members of thislibrary. Rather, smaller diagnostic sets of mutants wereto the opposing active site cavity (Figures 2A and 2B). In

STS, this loop is displaced relative to the corresponding constructed, expressed, purified, and assayed to quicklyisolate and identify the mechanistically relevant residues,loop in CHS (residues 132–137 in alfalfa CHS), ostensibly

due to the presence of proline (Figure 2B) at position as outlined below.This mutagenic approach revealed region B (area 4) to138 of STS (CHS position 135). Closer examination of

this buried region reveals compensatory mutations in be unimportant for cyclization specificity, but diagnosedregion A (areas 1–3, near the dimer interface) as criticaltwo other areas of primary sequence that in the folded

protein are juxtaposed either below (area 1) or above for mediating the Claisen to aldol cyclization switch inSTS. The product specificity of the area 4 mutant 0002(area 3) the displaced loop (area 2). These region A

differences in the newly solved STS structure did not is identical to wild-type CHS (data not shown), while theactivity of the 1210 mutant (8xCHS) closely resemblessignificantly alter the shape of the cyclization cavity pre-

viously identified in alfalfa CHS2 [12]. wild-type STS and the 2222 (18xCHS) mutant (Figure2E). The 1210 (8xCHS) mutant comprises five changes inConversely, region B is a 3 residue loop (area 4) lo-

cated on the outer, solvent-exposed surface of the STS the area 2 loop (Figure 2F), as well as one compensatorychange in area 1 (below the loop) and two compensatoryCoA binding tunnel (on the rim opposite the CoA phos-

phate binding residues) (Figure 2D), whose displace- changes in area 3 (above the loop) (Figure 2G). Unfortu-nately, nearly all of the more minimal mutants obtainedment relative to CHS was also observed in the only other

type III PKS crystal structure published to date, that of from this library (i.e., combinations 1200, 0210, 1110,1100, 0110, 1000, 0100, and 0010 from Figure 2C) exhib-a 2-pyrone synthase (2-PS) from daisy [16]. In 2-PS, this

displacement of region B residues toward the panteth- ited very little activity due to protein aggregation andprecipitation throughout the expression and purificationeine binding tunnel provides additional hydrogen bond-

ing and van der Waals contacts with bound CoA. We process. This latter result demonstrates the extent towhich the pine STS has structurally reinforced its func-were unable to obtain crystals of the STS-CoA complex.

Nevertheless, although the amino acid sequence of re- tionally important area 2 loop conformation since diverg-ing from its CHS ancestor. In the few cases where mu-gion B varies between these two enzymes, it seems

probable, as with 2-PS, that displacement of the STS tants more minimal than the STS-like 8xCHS werecatalytically active, their observed cyclization specificityregion B loop toward the CoA binding tunnel also causesfollowed the C6→C1 Claisen condensation route exhib-additional contacts with CoA, principally through theited by wild-type CHS (data not shown). Contrary toadenine moiety. This arrangement may enhance the on-expectations, none of the mutant enzymes from thisrate or slow the off-rate of CoA-linked starters, malonyl-library produced any significant amount of the de-CoA extenders, and reactive intermediates. This possi-railment tetraketide lactone product (pictured in Figureble kinetic effect could have mechanistic implications1D and discussed later) made in trace amounts by bothfor the balance of competing reactions occurring withinCHS and STS. This lack of significant derailment fromthe active site cavity.the wild-type cyclization pathways indicates a smooth,one-step transition from the C6→C1 Claisen to the

Mutagenic Conversions of Alfalfa Chalcone C2→C7 aldol condensation mechanism.Synthase into Functional Stilbene SynthasesTo determine whether any of these crystallographically Apo and Resveratrol-Complexed Crystalobserved structural differences correlate with the aldol Structures of Resveratrol-Producing 18xCHScyclization specificity of STS, we used two rounds of Two high-resolution structures of the 18xCHS mutant,

a functional stilbene synthase, were also determined bymutagenesis to convert regions A and B (areas 1–4) of


Figure 3. Thioesterase-like STS “Aldol Switch” Controls Cyclization Specificity

(A) Slightly different bound conformations of resveratrol observed in the complexed 18xCHS crystal structure (green and rose) correlate tomovements of the flexible Phe265 side chain, overlaid with the structure of the previously determined resveratrol [12] bound in wild-type CHS(light gray) and viewed down the CoA binding tunnel into the active site cavity. Positioning of resveratrol’s starter- and malonyl-derivedaromatic rings are similar to each other and to CHS-bound naringenin (shown in Figure 1B in a similar view).(B) C-� trace overlay of the displacement of the area 2 loop in STS (gold) and 18xCHS (green), compared to CHS (blue). Two orientationsillustrate the positions and movements of residues 131–133 (CHS numbering).(C) Stereoview of the 18xCHS STS-like “aldol switch” hydrogen bonds, showing the 1.9 A resolution 2Fo Fc electron density map (bluewirecage) contoured at 1 sigma.(D) “Aldol switch” hydrogen bonding differences (resulting from repositioning of the Thr132 side chain) in CHS-like and STS-like active sites,compared to each other and to the active site of thioesterase II (TEII) from E. coli ([18]; PDB code 1C8U). Distances incompatible with hydrogenbond formation are given in parentheses and indicated with double-headed arrows. Putative nucleophilic water positions are highlighted inyellow.(E) Thin layer chromatography (TLC) analysis of the cyclization specificities of mutants designed to disrupt the 18xCHS mutant’s aldol switchhydrogen bond network while preserving the 18xCHS STS-like conformational changes (see text).

protein X-ray crystallography, in the apo form (1.9 A) or conformational changes relative to the wild-type CHSstructure were observed in the 18xCHS mutant.with the resveratrol product of STS bound in the active

site cavity (2.1 A) (Table 1). Although a homology model Likewise, although the resveratrol-complexed crystalgrew in a different space group than the apo 18xCHSbased upon wild-type CHS, rather than STS, was used

as the search model for molecular replacement, the ini- crystal (see Table 1), no significant 18xCHS conforma-tional changes occurred upon binding of resveratrol.tial electron density maps and all subsequent maps re-

vealed STS-like conformational changes in the mutated Moreover, the position and orientation of resveratrol inthe active site of this functional stilbene synthase isregions (Figure 2D). Indeed, the conformations of these

regions in the alfalfa 18xCHS mutant’s area 2 and 4 loop nearly identical to the positions and orientations of bothresveratrol and naringenin (the flavanone resulting fromconformations were remarkably STS-like (18xCHS area

2 and 4 loops: C-� rmsd with STS � 0.35 A, and with the Michael addition isomerization of the unstable chal-cone product) previously observed in wild-type CHSCHS � 1.9 A). However, spotty electron density in the

18xCHS area 4 loop indicates this solvent-exposed loop crystal complexes [12] (Figure 3A). This latter result ar-gues against the likelihood of any drastic reorientationto be less conformationally restricted in the alfalfa

18xCHS mutant than in wild-type pine STS. No other within the STS active site cavity, relative to their posi-


tions in CHS, of either the coumaroyl starter molecule or ent at the back of all type III PKS active site cavities. Thisnewly identified STS hydrogen bonding configuration isthe subsequently formed linear polyketide intermediates

common to both enzymes. Two slightly different resver- similar, although not identical, to the catalytic machineryof a type II thioesterase recently discovered in E. coliatrol conformations in the four noncrystallographically

related monomers of the resveratrol-complexed 18xCHS [18] (Figure 3D). Significantly, a key difference betweenthe CHS and STS reactions is the need, in STS, forasymmetric unit (is equal to unit cell for P1 lattice) corre-

late with positional changes of the relatively disordered thioesterase activity to cleave the covalent bond linkingthe polyketide’s C1 carbon and the catalytic cysteine.Phe265 side chain (Figure 3A). Lack of order and posi-

tional differences are typical for this type III PKS “gate- In CHS, this bond is severed during the C6→C1 Claisencondensation reaction, but this same C1 thioester bondkeeper” residue [2, 12], and so the Phe265 differences

highlighted in Figure 3A should not be misinterpreted is not cleaved by STS’s C2→C7 aldol condensation reac-tion. Although the STS reaction pathway clearly requiresas unique or specific to this complexed structure.a thioesterase step, no such thioesterase-promotingresidues in STS have been identified or even proposed

An Aldol Cyclization Switch in Stilbene Synthase prior to this study. It should be noted that the kineticsalso Constitutes the Missing STS associated with thioesterase activity in the multifunc-Thioesterase Machinery tional STS active site cavity must be properly balancedWe next undertook a closer examination of the region A relative to polyketide initiation and chain elongation, leststructural differences implicated as functionally relevant these preceding steps be derailed by premature libera-by the 8xCHS mutant’s STS-like activity, in order to tion of thioester-linked polyketide intermediates fromprovide a sound mechanistic hypothesis for STS’s aldol the catalytic cysteine.cyclization specificity rather than a simple correlative The crystallographically observed thioesterase-likeexplanation based upon conformational differences. Di- configuration of residues in both STS and the 18xCHSvergent areas 1 and 3 are completely buried, and only mutant led us to hypothesize that the mechanistic ef-three residues from area 2 contact the type III PKS active fects of these observed changes in region A are notsite cavity. Residues 132 and 133 (alfalfa CHS number- steric, but rather mediated through the emergent STSing) border their own monomers’ active site cavity, adja- hydrogen bonding network. To test this novel hypothe-cent to the dyad-related monomer’s residue 137 side sis, we introduced subtle point mutations intended tochain. The 137 position, which is a methionine in alfalfa disrupt this hydrogen bonding system in the 18xCHSCHS and a leucine in Scots pine STS, is the only residue mutant (a functional stilbene synthase), while intention-in either enzyme that contributes to the active site cavity ally preserving the STS-like conformational changes weof the opposing monomer (Figure 1B). Prior to the eluci- had already introduced into alfalfa CHS. As predicted fordation of the STS structure, we had exchanged both electronic rather than steric control of aldol cyclizationenzymes’ position 137 amino acids based upon homol- specificity, each of these 18x(�1) CHS mutants (T132A,ogy modeling, but these mutations had no effect on S131A, and E192Q) exhibited increased chalcone pro-cyclization specificity or catalytic efficiency in either en- duction at the expense of stilbene production (Figurezyme (data not shown). Conversely, residues 132 and 3E). (Residue 131 is neither solvent exposed nor dis-133 (TT in CHS and TS in STS) were not previously placed by the area 2 “kink,” but its buried side chainsuspected to be important for aldol cyclization specific- hydroxyl also forms a hydrogen bond to the Glu192 sideity, but our comparison of the STS, CHS, and 18xCHS chain carboxylate.) While the E192Q mutation unfortu-mutant crystal structures reveals an important differ- nately causes enzyme instability (glutamate 192 is ap-ence caused by the displacement of the conserved parently also needed for fold stability, as it is absolutelyThr132 residue in the resveratrol-producing enzymes conserved among type III PKS enzymes), this mutant’s(Figure 3B). While the same buried conformational rear- cyclization specificity follows a CHS-like C6→C1 Claisenrangement of STS’s area 2 loop results in a more drastic condensation path. Conversely, the S131A and T132Alateral displacement of residue 133, it is unlikely that 18x(�1) CHS mutants are relatively active and stable,position 133 plays any significant steric or electronic and produce chalcone and resveratrol in ratios ofrole in the STS cyclization mechanism, due to its more roughly 50:50 and 75:25, respectively. Given the natureremote location in the active site cavity. and position of the 18x(�1) CHS T132A and S131A muta-

Conversely, the more subtle displacement of Thr132 tions, and considering the fold instability of other mutantcaused by the buried changes in the area 2 loop brings enzymes with fewer conformation-changing region Aits side chain hydroxyl moiety within hydrogen bonding mutations than those comprising the (1210) 8xCHS mu-distance of a Ser338-stabilized water molecule poised tant, the effect of these additional mutations is almostadjacent to the catalytic cysteine (Figures 3C–3D). It certainly electronic rather than steric. These resultsshould be noted that this active site water molecule is strongly indicate that STS aldol cyclization specificity issomewhat dynamic, as its position and occupancy can electronically mediated. Furthermore, aldol cyclizationdiffer in complexes and even between monomers in specificity is directly related to the observed thioester-the same crystal. However, this Ser338-stabilized water ase-like hydrogen bonding network connecting Glu192,molecule is crystallographically observed in the active through the repositioned Thr132 side chain hydroxyl, tosite cavities of both STS and CHS. In the STS-like active a water molecule poised next to the catalytic cysteine.site, the new hydrogen bond to Thr132 electronically Consequently, we have labeled the STS-like rearrange-connects this water molecule, through the repositioned ment of these residues the “aldol switch.”

Notably, the varying ratios of Claisen versus aldol cy-threonine hydroxyl, to a buried glutamate residue pres-


Figure 4. STS Mechanistic Options and Relevant Solution Chemistry

(A) Spontaneous solution-based polyketide C2→C7 aldol condensation cyclization chemistry leading to stilbenes. Atoms fated for eliminationas molecules of CO2 and H2O are colored in red and blue, respectively. The aromatized stilbene acid solution-based intermediate producthas been shown not to be an intermediate in the STS-catalyzed reaction (see text).(B) Plausible reaction pathways for the four STS cyclization-related events, assuming mechanistic divergence from CHS begins with an aldolswitch-catalyzed thioesterase-like hydrolytic step. Scenario One depicts a decarboxylative cyclization reaction, as described by Ebizuka’sgroup [14]. Scenario Two depicts two alternative decarboxylation schemes that follow a solution chemistry-like nondecarboxylative aldolcondensation-based cyclization. Atoms fated for elimination as molecules of CO2 and H2O are again colored red and blue, respectively.

clization specificity achieved with our 18x(�1) CHS universal STS consensus sequence. The only conservedtrend across these eight scattered residues in diverseT132A and S131A mutants represents a significant tech-

nological advance with implications for various PKS en- STSs is the substitution of a bulkier residue in place ofCHS’s Val98. However, we have quite recently solvedgineering projects, as it confirms that we can subtly

manipulate the type III PKS aldol switch region to direct the crystal structures of the peanut and grapevine STSenzymes, and our preliminary analysis of these STS en-the biosynthesis of desired mixtures of chalcone and

stilbene natural products emanating from a single mu- zymes demonstrates a conserved set of conformationalchanges in the aldol switch region, suggesting a con-tant enzyme (for example, in any plant species’ native

CHS enzyme, already optimized by natural selection for served STS mechanism despite this sequence variation.A more comprehensive functional and structural analy-expression in the cellular environment of that particular

species). Thus, engineering the aldol switch provides sis of the these peanut and grapevine STS aldol switchresidues is underway and will be published with thesea new biotechnological tool for optimizing resveratrol

production to confer antifungal and nutritional value in additional STS crystal structures in the near future.engineered plants without abrogating the essential pro-duction of chalcone-derived flavonoids. Toward an Overall Mechanism for the Multistep

Stilbene Synthase ReactionInterestingly, the eight pine STS-derived mutationsthat mechanistically convert alfalfa CHS to a stilbene As early as 1966, polyketide cyclization experiments

carried out in solution by Harris and others revealedsynthase are not conserved residues in the STS en-zymes from peanut or grapevine, confirming the lack of a that both Claisen and aldol intramolecular cyclization


specificities were readily achievable using aqueous con- events, only four of which begin, as our current resultssuggest, with hydrolysis of the C1 thioester bond.ditions over a range of pHs [19–22]. These elegant and

informative biomimetic studies found that the cycliza- Previously, the discovery of stilbenecarboxylic acidnatural products (stilbenes retaining the C1 carboxylatetion fate of linear polyketides is mediated in solution by

the presence or absence of an ester bond at the C1 at the C2 position) in a few plants (but notably not pine,peanut, or grape) fostered the parsimonious assumptioncarboxylate. More specifically, when C1 is part of an

ester (or thioester) bond, C6→C1 Claisen cyclization that stilbenecarboxylic acids must be on-pathway inter-mediates in stilbene biosynthesis (i.e., that STS-cata-predominates, while C2→C7 aldol cyclization is favored

when C1 exists as a free acid (Figure 4A). These early lyzed aldol cyclization precedes C1 decarboxylation)[13], like the solution chemistry reaction sequence de-findings foreshadow the most likely mechanistic inter-

pretation of our current structural and mutagenic results. picted in Figure 4A. Two recent pieces of evidence fromEbizuka’s group call into question this mechanistic as-In contrast to more recent proposals that have assumed

a steric mechanism of STS functional divergence from sumption [14]. First, a careful analysis of in vitro STSreaction products revealed absolutely no stilbenecar-CHS, our findings suggest that the cyclization fate of

the cysteine-linked tetraketide intermediate is deter- boxylic acid byproducts. Second, a deuterium-labelingexperiment established that stilbene decarboxylationmined by which of two competing processes occurs

first in the type III PKS active site cavity: cleavage of the precedes aromatization of the new ring. These factors,along with the observation that CHS is more likely thanC1-cysteine thioester bond by a CHS-like intramolecular

C6→C1 Claisen condensation, or hydrolysis of the same STS to produce lactone derailment products, promptedEbizuka’s group to conclude that the STS mechanismC1-cysteine thioester bond by an adjacent activated

nucleophilic water molecule to form a free carboxylic likely initiates by hydrolysis of the linear tetraketide’sC1 thioester linkage to cysteine, followed by a decarbox-acid intermediate. This acidic intermediate can then un-

dergo facile aldol cyclization in the active site. In other ylative aldol cyclization [14], as shown in Scenario Oneof Figure 4B.words, the kinetic balance between Claisen-mediated

ring cyclization and thioesterase-catalyzed hydrolysis The idea that aldol cyclization could be driven by anenergetically favorable elimination of C1 as a moleculeof the tetraketide intermediate results in partitioning be-

tween parallel mechanistic pathways. This latter pro- of CO2 is an appealing one. However, two lines of evi-dence from the early literature on biomimetic polyketidecess is upregulated in stilbene synthases by the emer-

gence of a thioesterase-like “aldol switch” hydrogen cyclization in solution suggest otherwise [19–22]. First,these studies indicate that C1 decarboxylation of linearbonding network that allows Glu192, through Thr132, to

activate a Ser338-positioned water molecule to a nu- polyketides competes with, rather than facilitates, theC2→C7 aldol cyclization of polyketides in solution. Fur-cleophilic hydroxide anion for hydrolysis through base

catalysis. thermore, these experiments also show that decarboxyl-ation does not accompany aldol cyclization in solution,From an evolutionary standpoint, obtaining an alterna-

tive cyclization fate by promoting thioester hydrolysis but can be thermally induced afterwards (Figure 4A).Since aldol cyclization of polyketides in solution doesto exploit intrinsic chemical reactivity is much simpler

than having to precisely reposition an identical tetrake- not require energetic coupling to decarboxylation, it fol-lows that there should be no energetic need to coupletide intermediate in an alternate productive conforma-

tion using steric reshaping of the type III PKS active STS’s enzymatic aldol cyclization reaction to decarbox-ylation, as implied by Scenario One.site cavity. Our current mechanistic proposal involving

kinetic partitioning between competing pathways is also Rather, our current structural and mutagenic results,when considered in light of solution-based polyketideconsistent with the observation that wild-type CHS and

wild-type STS each produce minor amounts (1%–5%) aldol cyclization studies, imply that the STS active siteis more likely to utilize a nondecarboxylative aldol cycli-of each other’s major reaction product in vitro [23, 24].

Although our elucidation of the structural and func- zation, as depicted in Scenario Two of Figure 4B. How-ever, the recent results from the Ebizuka group dictatetional basis for STS enzymes’ divergent aldol cyclization

specificity represents a significant achievement, as of that any STS mechanistic proposal that contains sucha nondecarboxylative cyclization must also include ayet we do not fully understand the complete STS reac-

tion mechanism. Five distinct reaction events accom- plausible mechanism for decarboxylating the resultingcyclized intermediate prior to ring aromatization. Logi-pany STS’s transformation of the linear tetraketide inter-

mediate into a stilbene scaffold: (1) hydrolytic cleavage cally, loss of the C1 carboxyl as CO2 could either occurprior to, or in conjunction with, the dehydrative elimina-of the C1 thioester bond to the catalytic cysteine, (2)

C2→C7 intramolecular aldol cyclization, (3) decarboxy- tion of the C7 hydroxyl group (Figure 4B). In fact, thereare plausible mechanistic options for both of theselative loss of C1 as CO2, (4) dehydrative elimination of

the tertiary alcohol derived from the C7 carbonyl oxygen, hypothetical postcyclization decarboxylation reactionpathways, which we will examine in greater detail below.and (5) aromatization of the new stilbene ring (Figure

4B). Since equilibrium favors the enol form of the C3 Beginning with a nondecarboxylative aldol cycliza-tion, both Scenario Two options utilize shielding (by theand C5 carbonyls, only the tetrahedral sp3 hybridization

of the predecarboxyled C2 and predehydrated C7 pre- exclusion of bulk solvent from the STS active site) toprevent the acid-catalyzed C7 dehydration reaction thatvents immediate postcyclization aromatization. For our

purposes, it is convenient to consider C7 dehydration produces aromatic stilbene acids in solution, as shownin Figure 4A, and instead employ intramolecular cyclicand ring aromatization as a single event, reducing the

number of variables from five to four. There are at least decarboxylation reactions as the next step after cycliza-tion. Presumably, active site shielding might also favoreight mechanistically plausible ways to order these four


Figure 5. Detailed Alternative STS Mechanistic Proposals

(A) Detailed mechanism for our proposed thioesterase-like aldol switch hydrolysis step and subsequent solution-like nondecarboxylative aldolcyclization. The putative hydrolytic water molecule is highlighted in yellow, and its atoms are colored red or green, to track their movement andultimate fates. Two enol tautomers of the proposed 2S,7R chiral intermediate are shown, as they are relevant to the alternative decarboxylationmechanisms presented in (B).(B) Two plausible alternative decarboxylation mechanisms (see text) involving intramolecular pericyclic electron movement (concerted orstepwise), proceeding from alternative enol tautomers of the cyclized intermediate shown in (A). Both decarboxylation alternatives requirethe enzymatic prevention of water-catalyzed dehydration leading to aromatized stilbene acids. Atom color denotes atoms derived from thecatalytic water molecule shown in (A). The coupled decarboxylation-dehydration mechanism proceeding from tautomer two returns the aldolswitch region to its precyclization state (dashed box, compare to [A], dashed box).(C) The proposed chiral intermediate modeled in the STS-like 18xCHS active site cavity (second panel), in comparison to the observed positionsof the putative hydrolytic water molecule and aldol switch residues (first panel). The position of the modeled chiral intermediate is based uponthe positions of observed resveratrol complexes and modeled precyclization polyketide intermediates. Observed hydrogen bonds are in green,and putative hydrogen bonds involving the modeled intermediate are rendered in blue. Preservation of the C7 hydroxyl bond to Thr132following cyclization requires a sterically permitted 30� rotation of the Thr132 side chain relative to its crystallographically observed depictedposition. The modeled intermediate’s chiral C2 and C7 positions are labeled with their predicted stereochemistry, and relevant protons onthe C1 carboxyl and C7 hydroxyl group are rendered in white, with their modeled orientation corresponding to that depicted for tautomer 2(see [A] and [B]).

these intramolecular decarboxylation mechanisms over lize proven solution chemistry where appropriate anddepart from solution chemistry in ways that make senseexternally catalyzed alternatives. In contrast to Scenario

One’s arbitrary decarboxylative C2→C7 cyclization in the context of the enclosed STS active site cavity.These more plausible mechanistic proposals embod-mechanism, both Scenario Two mechanistic options uti-


ied in Scenario Two are further elaborated in Figure competing linear �-ketoacid-like decarboxylation mecha-nism’s need for an additional and intermolecular acid-5. Following thioester hydrolysis by the aldol switch-

activated water molecule, the protonation of the C7 car- catalyzed dehydration step.Following the preceding cyclization step’s putativebonyl oxygen that accompanies the STS-mediated aldol

cyclization step may very well be catalyzed by the pres- proton transfer from Thr132 to the C7 oxygen moiety(Figure 5A), the position of the resulting C7 tertiary hy-ence of the same aldol switch hydrogen bonding net-

work, thus returning the aldol switch residues to their droxyl proton next to Thr132 in turn facilitates a hydro-gen bond between a lone pair of electrons on the C7active form through indirect abstraction (mediated by

Thr132) of the acidic proton on Glu192’s protonated hydroxyl oxygen and the C1 carboxyl’s acidic proton.This geometry favors the formation of a cyclic, intramo-carboxyl group (Figure 5A).

Like all polyketides, this cyclized intermediate is capa- lecular six-membered transition state that simultane-ously promotes proton transfer from the C1 carboxylicble of achieving several different combinations of keto-

enol tautomers. In Figure 5B, each intramolecular decar- acid to the C7 hydroxyl, elimination of these respectivering substituents as molecules of CO2 and H2O, andboxylation reaction is depicted as proceeding from the

tautomeric form of the cyclized polyketide that would formation of a double bond between C2 and C7.While our proposed dehydrative decarboxylationseem to most favor that reaction. Both alternative decar-

boxylation reactions can exploit a six-center transition mechanism has not been described to date, this intra-molecular elimination (Ei) mechanism closely resemblesstate arising from the intramolecular movement of three

electron pairs. Our depiction here of cyclic intramolecu- a number of documented mechanistic precedents, al-though in solution this elimination mechanism usuallylar electron movement as concerted is not an assertion

of simultaneous electron movement. Rather, we merely requires thermal energy to achieve the relatively unfa-vorable cis-periplanar conformation necessary for theintend to convey that these six-centered intramolecular

reactions have the potential to occur simultaneously, reaction [28–31]. In fact, these intramolecular eliminationreactions most often occur in molecules where sigmabypassing the production of high-energy polar interme-

diates [25]. bond rotation to achieve the lower energy anti-peripla-nar alternative productive conformation is sterically pre-One Scenario Two intramolecular decarboxylation op-

tion, suggested to us by a colleague (G. Weiss, personal vented by bulky substituents or by incorporation in ringsystems [25]. In relation to the STS mechanism, we ear-communication), mirrors the concerted, six-center de-

carboxylation reaction of linear �-keto acids [26, 27] lier stated that shielding in the active site cavity likelyprevents the competing C7 dehydration reaction that(Figure 5B). In this mechanistic scenario, the C3-derived

carbonyl oxygen both abstracts the acidic proton from produces stilbene acids following solution-based aldolcyclization reactions. This blocked dehydration reactionthe C1 carboxyl moiety and also acts as an electron sink

for decarboxylation of the C1 carboxyl group. Following undoubtedly proceeds in solution via the lower energyanti-periplanar elimination of the remaining C2 proton.this intramolecular reaction, a distinct acid-catalyzed

protonation of the C7 hydroxyl group would be required Although the documented Ei reaction precedents em-ploy seemingly more suitable leaving groups than theto initiate C7 dehydration, presumably accompanied by

aromatization of the stilbene’s new resorcinol ring. Since C7 tertiary alcohol depicted here, the combination ofthe very favorable intramolecular acid-catalyzed proton-STS does not produce stilbene acid derailment prod-

ucts, any such STS-catalyzed dehydration activity ation of this tertiary alcohol and the aromatization of theresveratrol product may serve to make the resultingwould need to scrupulously discriminate between the

initial carboxyl-bearing and subsequently decarboxyl- water molecule an equally favorable leaving group.While these alternative Scenario Two intramolecularated cyclized intermediates. Moreover, it is not obvious

from inspection of the STS structure what active site decarboxylation mechanistic proposals require verysimilar catalytic requirements from the STS active sitefeature would catalyze this dehydration function. Dehy-

dration of this putative decarboxylated intermediate to (i.e., shielding), they differ in their stereochemical re-quirements. Although the aromatic stilbene product isproduce the lower energy aromatic stilbene scaffold of

resveratrol seems more likely to occur in solution, fol- achiral, the initial cyclized intermediate produced by anondecarboxylative aldol condensation possesses twolowing diffusion out of the STS active site cavity. Pre-

sumably, the thermal C2 decarboxylation of C7-dehy- chiral carbons (C2 and C7). While the �-ketoacid decar-boxylation proposal is consistent with any of these ste-drated stilbene acids in solution (Figure 4A) proceeds

by a mechanism similar to this first proposal for C2 reoisomers, the coupled decarboxylation-dehydrationproposal requires the initial cyclized intermediate todecarboxylation by STS.

The other plausible Scenario Two intramolecular de- have a cis arrangement of its C7 hydroxyl and C1 car-boxyl groups, and thus is consistent with only two ste-carboxylation mechanism, inspired by three-dimen-

sional modeling of intermediates in the STS active site, reoisomers (2S, 7R or 2R, 7S ). Significantly, this cisjuxtaposition of the C7 hydroxyl group and the C1 car-utilizes a similar intramolecular cyclic transition state at

the C2 carboxyl position, but one which involves an boxyl moiety, which allows the C1 free acid’s acidicproton to catalyze the intramolecular dehydration of theinteraction with the C7 hydroxyl moiety rather than the

C3 carbonyl group (Figure 5B, coupled decarboxylation/ C7 hydroxyl, is energetically favored in these two stereo-isomers because of the favorable orientation of the C1dehydration option). This alternative six-center reaction

immediately achieves the lower energy aromatic stil- carboxyl on C2 trans to C7’s other substituent (the muchlarger starter-derived coumaroyl moiety).bene scaffold of resveratrol by coupling C1 decarboxyl-

ation, intramolecular acid-catalyzed C7 hydroxyl loss, Based upon the three-dimensional modeling of bothlinear and cyclized intermediates in the STS active siteand aromatization of the stilbene ring, bypassing the


cavity, we predict that the initial cyclized intermediateresulting from a nondecarboxylative C2→C7 aldol cycli-zation in the STS active site cavity is most likely topossess 2S, 7R stereochemistry. Figure 4C models thispredicted intermediate in the STS active site cavity. Theposition shown is based on the observed position andorientation of bound resveratrol (Figure 3A), the locationof the putative hydrolytic water molecule, and the mod-eled positions and orientations of precyclization tetrake-tide intermediates.

The putative 2S, 7R stereochemistry of the cyclizedintermediate represents the first stereochemical pre-diction of a transitory chiral STS reaction intermediate.Although this stereochemistry is compatible with bothof our postcyclization intramolecular decarboxylationmechanistic proposals, our structure-based predictionof this cyclized intermediate’s stereochemistry may fa-cilitate the design of potential STS reaction intermediateanalogs or inhibitors that can be used to distinguish onedecarboxylation mechanism over the other. Similarly, avery recent analysis found the catalytic cysteines of CHSand STS enzymes to differ significantly from each otherin their susceptibility to inhibition by various chloroacet-amide herbicides [32]. This result was unexpected, asthe relative susceptibility of condensing enzymes to par-ticular inhibitors most often correlates with differencesin their active site volumes. In retrospect, the differentresponses of CHS and STS enzymes to these chloro-acetamide herbicides may be mediated by the aldolswitch region differences we report here, given the closeproximity of the catalytic cysteine and the aldol switchregion.

Insights into Stilbenecarboxylic Acid BiosynthesisIt is fitting to conclude this analysis with some structuraland mechanistic insights into the related topic of stil-benecarboxylate (also known as stilbenecarboxylic acidor stilbene acid) biosynthesis. Like stilbenes, these rare

Figure 6. Stilbenecarboxylic Acid Biosynthetic Proposalnatural products have been isolated from a small but

(A) Biological stilbene acids discussed in the text (hydrangic anddiverse collection of plants, including species of Hydran- lunularic acids). Note missing hydroxyls (compared to generic stil-gea and primitive liverworts of the Marchantia family [2]. bene acid in Figure 1D or in [B]) indicating reduction of a polyketideLabeling experiments and the incorporation of stilbene- intermediate.

(B) Novel proposal for stilbene acid biosynthesis based upon solu-carboxylates into well-characterized downstream natu-tion chemistry (see Figure 4A and text) and observed product speci-ral products demonstrate that the isolation of in vivoficities of CTAS enzymes (see text). O5→C1 lactonization reactionsstilbene acids is neither an artifact nor a by-productare shown in green, with C2→C7 aldol condensations shown in red.

of stilbene biosynthesis [14]. These natural productspresumably result via a nondecarboxylative STS-like(C2→C7 aldol) type III PKS reaction mechanism. The acid lactone (CTAL), a minor tetraketide CHS derailment

product, from p-coumaroyl-CoA and three molecules ofsame reaction pathway, but proceeding from an ali-phatic hexanoyl starter, is likely to be responsible for malonyl-CoA [33]. This enzyme was named p-coumaro-

yltriacetic acid synthase (CTAS) (Figures 1D and 6B).the carboxylate-bearing resorcinol moiety of tetrahydro-cannabinol (THC) compounds in Cannabis. While candi- Schroder’s group independently observed the same lac-

tone product specificity with the only apparent non-date type III PKS enzymes have been cloned from somespecies, in vitro reconstitution of stilbenecarboxylate CHS type III PKS in Hydrangea macrophylla L. (Garden

Hortensia), a species containing both hydrangic acidbiosynthesis has been problematic at best.Two independent attempts to clone and characterize and the similar lunularic acid (Figure 6A) [34]. However,

when primed with dihydro-p-coumaroyl-CoA as a starter,stilbenecarboxylic acid synthases have been published,with each group finding only one candidate (i.e., non- this enzyme produced nearly 50% 5-hydroxylunularic

acid, a stilbenecarboxylic acid, in addition to triketideCHS) type III PKS enzyme [33, 34]. Ebizuka’s group,searching for Hydrangea macrophylla var. thunbergii’s and tetraketide lactones. Notably, this group also de-

tected 5-hydroxylunularic acid when a pine CHS washydrangic acid (Figure 6A) synthase, cloned an enzymethat in vitro efficiently produced p-coumaroyltriacetic given the dihydro-p-coumaroyl-CoA starter, while a pine


STS given the same starter failed to produce a stilbene- studies demonstrate that they undergo spontaneous al-dol cyclization when placed in basic or mildly acidiccarboxylic acid. These results are extremely puzzling

if one assumes stilbenes and stilbenecarboxylates are conditions [19–22]. Furthermore, while triketide lactonesare quite stable in solution, tetraketide lactones, overproduced by a conserved biosynthetic mechanism.

Both publications discussed plausible explanations time, are much more likely to reopen, and in neutral orbasic solution, the resulting linear free acid is likely tofor these expected stilbenecarboxylate synthases’ sur-

prising in vitro activities, including the influence of a undergo C2→C7 aldol cyclization rather than relactoni-zation [20]. Thus, even if the observed Hydrangea PKSmissing polyketide ketoreduction step (reviewed in [2])

indicated by the hydroxylation pattern of the Hydrangea lactones form in the active site, the tetraketide-derivedCTAL is considerably less stable in solution than is thestilbenecarboxylates (see Figure 6). Schroder’s group

noted that ketoreduction of the C5 polyketide position triketide-derived bis-noryangonin, and thus much morelikely under physiological conditions to reopen and sub-or use of the dihydro-p-coumaroyl starter both interrupt

the extensive bond conjugation system possible in sequently undergo spontaneous aldol cyclization toform a stilbenecarboxylate. Unlike STS’s type III PKS-the unreduced p-coumaroyl-derived tetraketide inter-

mediate [34]. catalyzed aldol cyclization, nonenzymatic C2→C7 aldolcyclizations are almost never accompanied by sponta-However, now that the structural basis for STS-like

aldol cyclization specificity has been elucidated, it is neous decarboxylation, overwhelmingly producing stil-benecarboxylates as their stable end products. In fact,clear that the Hydrangea CTAS sequences contain no

such STS-like signatures and are in fact quite CHS- the decarboxylation of stilbenecarboxylates in solutionrequires considerable thermal energy [22].like aside from one important (but non-STS-like) T197N

active site substitution. Notably, CHS enzymes produce Consistent with the Schroder group’s results, the na-ture of the R group attached to a linear tetraketide alsomore in vitro lactone derailment products than do STS

enzymes, even when provided with their physiological modulates its rate of spontaneous aldol cyclization. Inthe linear tetraketide intermediates formed by type IIIstarters [23]. Although lactone derailment products are

often assumed to form in solution after hydrolytic re- PKS enzymes, this R group represents the entire startermolecule-derived portion except for the C7 ketone (seelease of linear free acid polyketide intermediates, our

current study of STS suggests hydrolysis in the STS R� in Figure 1C). While a methyl R group facilitates veryrapid solution aldol cyclization, other substituents pro-active site cavity leads to the production of stilbenes.

It therefore seems likely that lactone formation initiates mote slower processes, especially those containing a bond conjugated with the C7 ketone [22, 35], such asCTAS-like product off-loading via an intramolecular at-

tack by the C5 keto-enol oxygen on the cysteine-teth- in p-coumaroyl-derived (but not dihydro-p-coumaroyl-derived) tetraketides. Notably, all of the Hydrangea PKSered C1 thioester. Similar to CHS’s C6→C1 cyclization

and product release, and distinct from STS’s water- in vitro assays were extracted after 30–60 minutes [33,34], which is too short of a time frame to allow for mostmediated thioester hydrolysis, the C1-tethered tetrake-

tide conformation leading to CTAL is likely favored by spontaneous aldol cyclizations to occur [22, 35]. Inlight of these facts, the supposedly enzymatic in vitrothe position 197 substitution. Significantly, a T197L point

mutant of the well-characterized alfalfa CHS was pre- formation of stilbenecarboxylates only from a dihydro-p-coumaroyl-CoA starter, especially when associatedviously shown to mirror CTAS’s fairly exclusive produc-

tion of CTAL [16]. with high yields of tri- and tetraketide lactones [34], ismore likely to result from spontaneous solution aldolThe biomimetic polyketide cyclization literature dating

back to the late 1960’s reveals several pieces of circum- cyclization of a linear tetraketide (reopened lactone).Given a longer reaction time before acid quenching, thestantial evidence that together suggest a new interpreta-

tion of both the in vitro activity of CTAS enzymes and p-coumaroyl-initiated CTAS reaction would likely alsoresult in the spontaneous and very favorable productionthe in vivo production of stilbenecarboxylates that has

not been addressed in the CTAS literature. These stud- of stilbenecarboxylates.ies suggest that the cloned Hydrangea CTAS enzymesare in fact responsible for the in vivo biosynthesis of the

Significancestilbenecarboxylic acid precursor CTAL. Most notably,in solution the labile CTAL lactone can precede the

Chalcone synthase (CHS) and stilbene synthase (STS)spontaneous aldol condensation leading to stilbenecar-are related type III polyketide synthase (PKS) enzymesboxylic acids. We propose that the final aldol cyclizationthat catalyze the formation of identical linear tetrake-step is not catalyzed by a type III PKS at all, and is likelytide intermediates. However, each catalyzes the cycli-to have emerged in Hydrangea and other stilbenecar-zation of this reactive polyketide intermediate usingboxylate-producing species as the natural consequencealternative intramolecular condensation mechanisms.of forming tetraketide lactones at physiological pH. Sig-CHS, ubiquitous in the plant kingdom, catalyzes anificantly, the standard assays employed in the pub-C6→C1 Claisen condensation to form the core chal-lished CTAS studies would prevent detection of all butcone scaffold of all flavonoid natural products. Con-the most facile solution-phase aldol cyclizations. It isversely, STS has evolved in a limited number of phylo-also possible that some unidentified enzyme may subse-genetically distinct plants via gene duplication andquently have been recruited to catalyze this spontane-subsequent mechanistic divergence from CHS, andous but slow process.instead catalyzes a C2→C7 aldol condensation thatWhile free �-keto carboxylic acids (i.e., linear polyke-

tides) readily form lactones in acidic solution, solution forms the stilbene backbone of resveratrol and related


crystallization buffer, in the presence of 1 mm DTT. The wild-typeantifungal phytoalexins. Stilbenes can confer antifun-STS crystallization buffer contained 13% (w/v) PEG 8000, 300 mMgal resistance to host plants when STS is heterolo-ammonium acetate, and 100 mM MOPSONa� buffer at pH 7.0.gously expressed in them, and as phytonutrients inPrior to freezing in liquid nitrogen, STS crystals were equilibrated

mammalian diets, they also possess a number of for 2 min in a cryogenic buffer identical to the crystallization bufferhealth benefits for humans and other animals. Here we except for the use of 15% (w/v) PEG 8000 and the inclusion of 10%

(v/v) glycerol, and then moved into a similar solution containing 25%present the first STS crystal structure and the detailed(v/v) glycerol for another 2 min. The apo 18xCHS crystallizationstructural comparison to previously characterizedbuffer contained 21% (w/v) PEG 8000, 300 mM ammonium acetate,CHS that guided our subsequent mutagenic conver-100 mM HEPESNa� buffer at pH 7.5, and 3% (v/v) ethylene glycol,sion of alfalfa CHS into a functional STS. Significantly,whereas the 18xCHS resveratrol complex crystal was obtained by

we identify the previously obscure structural basis for the inclusion of 2.5 mM trans-resveratrol in the 18xCHS crystalliza-the evolution of STSs from their CHS ancestors. Unex- tion buffer. Prior to freezing, both the apo and resveratrol-com-

plexed 18xCHS crystals were subjected to a 60 s cryogenic soakpectedly, the mechanism of STS functional divergencein crystallization buffer modified to contain 23% (w/v) PEG 8000stems from the emergence of a cryptic thioesteraseand 6% (v/v) ethylene glycol.activity in the active site, due to an alternative hydro-

P. sylvestris STS crystallized in the P21 space group, with unit cellgen bonding network termed the “aldol switch.” Thisdimensions of a � 57.2 A, b � 361.3 A, c � 57.3 A, � � � � 90.0�,

mechanism is distinct from previous models of type and � � 98.4�, and six monomers (three physiological homodimers)III PKS functional divergence that imply that steric in the asymmetric unit. The apo 18xCHS crystals also grew in the

P21 space group, but with a much smaller unit cell of a � 71.6 A,factors that shape the active site cavity thereby directb � 59.8 A, c � 82.5 A, � � � � 90.0�, and � � 108.2�, with onlythe conformation and thus the cyclization fate of poly-two monomers (one homodimer) in the asymmetric unit. Conversely,ketide intermediates. Finally, we propose a novelthe 18xCHS resveratrol complex crystallized in the P1 space group,mechanistic hypothesis, derived from existing biomi-with unit cell dimensions of a � 64.3 A, b � 71.7 A, c � 85.7 A, � �

metic polyketide cyclization studies, that explains a 111.4�, � � 91.6�, and � � 90.1�, and four monomers (two dimers)series of puzzling experimental results in the related in the unit cell (asymmetric unit).

Data were collected at the Stanford Synchrotron Radiation Labo-field of stilbenecarboxylic acid biosynthesis. The dis-ratory (SSRL) or at the European Synchrotron Radiation Facilitycovery of the structural basis of stilbene biosynthesis(ESRF). Diffraction images were indexed and integrated with DENZOfills a major gap in our understanding of the structural[37], the reflections were merged with SCALEPACK [37], and dataand mechanistic underpinnings of functional diversityreduction was completed with CCP4 programs [38] or using XDS [39].

in the type III PKS enzyme superfamily, as aldol con- Despite strong diffraction, overlapping reflections caused by onedensation-driven polyketide cyclization was one of the unusually long STS unit cell axis (b � 361.3 A) seriously complicated

this particular data collection. Although the application of an addi-least understood type III PKS reactions. The success-tional spacial deconvolution step carried out using PROW [40] im-ful structurally guided engineering of efficient STS ac-proved completeness (without increasing the Rsym), many problematictivity into CHS demonstrates the utility of this com-overlapping reflections remained, resulting in only 74% complete-bined structure-function approach in numerous typeness and less than ideal crystallographic statistics (see Table 1).

III PKS engineering projects.

Structure Determination and RefinementExperimental Procedures All structures were solved by molecular replacement using EPMR

[41] or AMoRe [42]. For both the wild-type STS and the apo 18xCHSMutagenesis structures, the search model consisted of a homology model basedSingle or a minimal set of multiple site mutations in contiguous areas upon the alfalfa CHS2 crystal structure, generated with MODELLERof primary sequence were introduced using the QuikChange system [43]. The refined apo 18xCHS structural model was used as the(Stratagene). More extensive mutations in each area were generated search model for the 18xCHS resveratrol complex.using a second round of QuickChange with new mutagenic primers Solutions were iteratively refined using CNS [44] or REFMAC [45].and a CHS gene template already mutated to the “intermediate” Inspection of the |2Fo Fc| and |Fo Fc| electron density mapssequence shown in Figure 2C. Mutagenesis at multiple distal sites and model building were performed in O [46]. The final refinementwas achieved by PCR amplification of wild-type (or “intermediate” statistics for each structure are listed in Table 1. Each residue’smutant) regions using 5� and 3� primers containing mutagenic backbone conformation was categorized (by CCP4’s PROCHECKchanges. Following purification using agarose gel electrophoresis, analysis of Ramachandran plots [38]) as either core (most favorable),these overlapping fragments were assembled and amplified by PCR allowed, generally allowed, or disallowed. The percentage of thewith the appropriate end primers to generate contiguous mutant final pine STS model’s residues in each group is 87.9%, 11.2%,megaprimers that spanned several distal sites and contained all of 0.8%, and 0.05%, respectively. The corresponding values for thethe desired mutations. These megaprimers were then separately 18xCHS apo structure’s residues are 90.3%, 8.8%, 0.75%, and 0.15%,incorporated into the full-length alfalfa CHS gene (in the pHIS-8 while the results with the resveratrol-complexed 18xCHS structure’sexpression vector construct [36]) by the QuickChange method. Each residues are 90.5%, 9.2%, 0.3%, and 0.075%. Each structure hadmutation was confirmed by automated nucleotide sequencing (Salk only one residue (of only one monomer) in a disallowed conforma-Institute DNA sequencing facility). tion, in each case a glutamine (position 234 in STS, 231 in CHS)

involved in a hairpin turn at the protein surface (distant from theProtein Expression and Purification active site). Notably, a similar backbone conformation of Gln231P. sylvestris STS was subcloned into the pHIS-8 expression vector, was observed in the previously reported wild-type CHS struc-as previously described for M. sativa CHS2 [36]. Following overex- ture [12].pression in E. coli BL21(DE3) cells, recombinant proteins were puri- Structural illustrations were prepared with MOLSCRIPT [47] andfied to homogeneity, concentrated to between 5 and 50 mg/ml, and were rendered with POV-Ray [48].stored at 80�C following buffer exchange into 12 mM HEPES (pH7.5), 25 mM NaCl, and 5 mM DTT, as described previously [36]. Enzyme Assays and Determination of Kinetic Constants

Mutant CHS enzyme assays were conducted as detailed elsewhere[36], in this case side-by-side with both wild-type CHS and wild-Crystallization and Data Collection

P. sylvestris STS and the 18xCHS mutant were crystallized by vapor type STS enzymes, all utilizing p-coumaroyl-CoA and [2-C14]-malo-nyl-CoA substrates and including a final reaction mixture acidifica-diffusion in hanging drops consisting of a 1:1 mixture of protein and


tion step to maximize detection of any potential lactone derailment chalcones, and 6�-deoxychalcones. J. Biol. Chem. 270, 7922–7928.products. Radiolabeled products were extracted, separated by TLC

as described prevously [36], and then visualized with film or with a 12. Ferrer, J.L., Jez, J.M., Bowman, M.E., Dixon, R.A., and Noel,J.P. (1999). Structure of chalcone synthase and the molecularMolecular Dynamics PhosphorImager system. Product ratios were

quantified using Molecular Dynamics ImageQuant software. basis of plant polyketide biosynthesis. Nat. Struct. Biol. 6,775–784.Steady-state kinetic constants were determined from linear initial

velocity measurements utilizing radiolabeled malonyl-CoA, Ecolume 13. Schroder, J. (1997). A family of plant-specific polyketide syn-thases: facts and predictictions. Trends Plant Sci. 2, 373–378.scintillation fluid, and a scintillation counter, as detailed previ-

ously [36]. 14. Shibuya, M., Nishioka, M., Sankawa, U., and Ebizuka, Y. (2002).Incorporation of three deuterium atoms excludes intermediacyof stilbenecarboxylic acid in stilbene synthase reaction. Tetra-Acknowledgmentshedron Lett. 43, 5071–5074.

15. Schanz, S., Schroder, G., and Schroder, J. (1992). Stilbene syn-M.B.A. thanks J. Jez for introduction to type III PKS radiochemicalthase from Scots pine (Pinus sylvestris). FEBS Lett. 313, 71–74.assays, kinetics, and crystallographic methods. We thank B. Moore,

16. Jez, J.M., Austin, M.B., Ferrer, J., Bowman, M.E., Schroder, J.,M. Burkhart, and G. Weiss for helpful mechanistic discussions. Thisand Noel, J.P. (2000). Structural control of polyketide formationmaterial is based upon work supported by the National Sciencein plant-specific polyketide synthases. Chem. Biol. 7, 919–930.Foundation under Grant No. 0236027 to J.P.N. Portions of this re-

17. Schroder, J. (1999). The chalcone/stilbene synthase-type familysearch were carried out at the Stanford Synchrotron Radiation Labo-of condensing enzymes. In Comprehensive Natural Reports,ratory, a national user facility operated by Stanford University onD. Barton, K. Nakanishi, O. Meth-Cohn, and U. Sankawa. eds.behalf of the U.S. Department of Energy, Office of Basic Energy(Oxford, UK: Elsevier). pp. 749–772.Sciences. The SSRL Structural Molecular Biology Program is sup-

18. Li, J., Derewenda, U., Dauter, Z., Smith, S., and Derewenda,ported by the Department of Energy, Office of Biological and Envi-Z.S. (2000). Crystal structure of the Escherichia coli thioesteraseronmental Research, and by the National Institutes of Health, Na-II, a homolog of the human Nef binding enzyme. Nat. Struct.tional Center for Research Resources, Biomedical TechnologyBiol. 7, 555–559.Program. We also acknowledge the European Synchrotron Radia-

19. Harris, T.M., and Carney, R.L. (1966). Biogenetically modeledtion Facility for provision of synchrotron radiation facilities. The tech-synthesis of �-resorcylic acids. J. Am. Chem. Soc. 88, 2053–nology associated with this manuscript has been licensed through2054.the Salk Institute to Allylix, Inc. The corresponding author (J.P.N.)

20. Bentley, R., and Zwitkowits, P.M. (1967). Biosynthesis of tropo-is a cofounder of Allylix, Inc.lones in Penicillium stipitatum. VII. The formation of polyketidelactones and other nontropolone compounds as a result of ethi-Received: February 26, 2004onine inhibition. J. Am. Chem. Soc. 89, 676–680.Revised: May 6, 2004

21. Harris, T.M., and Carney, R.L. (1967). Synthesis of 3,5,7-triketoAccepted: May 19, 2004acids and esters and their cyclizations to resorcinol and phloro-Published: September 17, 2004glucinol derivatives. Models of biosynthesis of phenolic com-pounds. J. Am. Chem. Soc. 89, 6734–6741.

References 22. Harris, T.M., and Harris, C.M. (1986). Biomimetic syntheses ofaromatic polyketide metabolites. Pure Appl. Chem. 58, 283–294.1. Schroder, J. (2000). The family of chalcone synthase-related

23. Yamaguchi, T., Kurosaki, F., Suh, D.Y., Sankawa, U., Nishioka,proteins: functional diversity and evolution. Recent Adv. Phyto-M., Akiyama, T., Shibuya, M., and Ebizuka, Y. (1999). Cross-chem. 34, 55–89.reaction of chalcone synthase and stilbene synthase overex-2. Austin, M.B., and Noel, J.P. (2003). The chalcone synthase su-pressed in Escherichia coli. FEBS Lett. 460, 457–461.perfamily of type III polyketide synthases. Nat. Prod. Rep. 20,

24. Suh, D.Y., Kagami, J., Fukuma, K., Iwanami, N., Yamazaki, Y.,79–110.Yurimoto, H., Sakai, Y., Kato, N., Shibuya, M., Ebizuka, Y., et3. Schroder, J., and Schroder, G. (1990). Stilbene and chalconeal. (2000). Chalcone and stilbene synthases expressed in eu-synthases: related enzymes with key functions in plant-specificcaryotes exhibit reduced cross-reactivity in vitro. Chem. Pharm.pathways. Z. Naturforsch. [C] 45, 1–8.Bull. (Tokyo) 48, 1051–1054.4. Tropf, S., Lanz, T., Rensing, S.A., Schroder, J., and Schroder,

25. Carey, F.A., and Sundberg, R.J. (1990). Advanced OrganicG. (1994). Evidence that stilbene synthases have developedChemistry. Part B: Reactions and Synthesis. Third Edition (Newfrom chalcone synthases several times in the course of evolu-York: Plenum Press).tion. J. Mol. Evol. 38, 610–618.

26. Jencks, W.P. (1969). Catalysis in Chemistry and Enzymology.5. Hain, R., Reif, H.J., Krause, E., Langebartels, R., Kindl, H., Vor-(New York: McGraw-Hill).nam, B., Wiese, W., Schmelzer, E., Schreier, P.H., Stocker, R.H.,

27. Westheimer, F.H., and Jones, W.A. (1941). The effect of solventet al. (1993). Disease resistance results from foreign phytoalexinon some reaction rates. J. Am. Chem. Soc. 63, 3283–3286.expression in a novel plant. Nature 361, 153–156.

28. Cope, A.C., and Trumbull, E.R. (1960). Olefins from amines: the6. Hipskind, J.D., and Paiva, N.L. (2000). Constitutive accumulationHofmann elimination reaction and amine oxide pyrolysis. Org.of a resveratrol-glucoside in transgenic alfalfa increases resis-React. 11, 317–493.tance to Phoma medicaginis. Mol. Plant Microbe Interact. 13,

29. Clive, D.L.J. (1978). Modern organoselenium chemistry. Tetra-551–562.hedron 34, 1049–1132.7. Siemann, E.H., and Creasy, L.L. (1992). Concentration of phyto-

30. DePuy, C.H., and King, R.W. (1960). Pyrolytic cis eliminations.alexin resveratrol in wine. Am J Enol Vitic. 43, 49–52.Chem. Rev. 60, 431–457.8. Soleas, G.J., Diamandis, E.P., and Goldberg, D.M. (1997). Res-

31. Nace, H.R. (1962). The preparation of olefins by the pyrolysisveratrol: a molecule whose time has come? And gone? Clin.of xanthates. The Chugaev reaction. Org. React. 12, 57–100.Biochem. 30, 91–113.

32. Eckermann, C., Matthes, B., Nimtz, M., Reiser, B., Lederer, B.,9. Fremont, L. (2000). Biological effects of resveratrol. Life Sci. 66,Boger, P., and Schroder, J. (2003). Covalent binding of chloro-663–673.acetamide herbicides to the active site cysteine of plant type10. Howitz, K.T., Bitterman, K.J., Cohen, H.Y., Lamming, D.W., Lavu,III polyketide synthases. Phytochemistry 64, 1045–1054.S., Wood, J.G., Zipkin, R.E., Chung, P., Kisielewski, A., Zhang,

33. Akiyama, T., Shibuya, M., Liu, H.M., and Ebizuka, Y. (1999).L.L., et al. (2003). Small molecule activators of sirtuins extendp-Coumaroyltriacetic acid synthase, a new homologue of chal-Saccharomyces cerevisiae lifespan. Nature 425, 191–196.cone synthase, from Hydrangea macrophylla var. thunbergii.11. Tropf, S., Karcher, B., Schroder, G., and Schroder, J. (1995).Eur. J. Biochem. 263, 834–839.Reaction mechanisms of homodimeric plant polyketide syn-

34. Eckermann, C., Schroder, G., Eckermann, S., Strack, D.,thase (stilbenes and chalcone synthase). A single active site forthe condensing reaction is sufficient for synthesis of stilbenes, Schmidt, J., Schneider, B., and Schroder, J. (2003). Stilbenecar-


boxylate biosynthesis: a new function in the family of chalconesynthase-related proteins. Phytochemistry 62, 271–286.

35. Howarth, T.T., Murphy, G.P., and Harris, T.M. (1969). Preparationand biogenetic-type aromatizations of tetraacetic acid (3,5,7-trioxooctanoic acid). J. Am. Chem. Soc. 91, 517–518.

36. Jez, J.M., Ferrer, J.L., Bowman, M.E., Dixon, R.A., and Noel,J.P. (2000). Dissection of malonyl-coenzyme A decarboxylationfrom polyketide formation in the reaction mechanism of a plantpolyketide synthase. Biochemistry 39, 890–902.

37. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray dif-fraction data collected in oscillation mode. Methods Enzymol.276, 307–326.

38. Dodson, E.J., Winn, M., and Ralph, A. (1997). Collaborative Com-putational Project, Number 4: providing programs for proteincrystallography. Methods Enzymol. 277, 620–633.

39. Kabsch, W. (1993). Automatic processing of rotation diffractiondata from crystals of initially unknown symmetry and cell con-stants. J. Appl. Crystallogr. 26, 795–800.

40. Bourgeois, D., Nurizzo, D., Kahn, R., and Cambillau, C. (1998).An integration routine based on profile fitting with optimizedfitting area for the evaluation of weak and/or overlapped two-dimensional laue or monochromatic patterns. J. Appl. Crys-tallogr. 31, 22–35.

41. Kissinger, C.R., Gehlaar, D.K., and Fogel, D.B. (1999). Rapidautomated molecular replacement by evolutionary search. ActaCrystallogr. D Biol. Crystallogr. 55, 484–491.

42. Navaza, J. (2001). Implementation of molecular replacement inAMoRe. Acta Crystallogr. D Biol. Crystallogr. 57, 1367–1372.

43. Sali, A., and Blundell, T.L. (1993). Comparative protein modelingby satisfaction of spatial restraints. J. Mol. Biol. 234, 779–815.

44. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros,P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges,M., Pannu, N.S., et al. (1998). Crystallography and NMR system:a new software suite for macromolecular structure determina-tion. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921.

45. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refine-ment of macromolecular structures by the maximum-likelihoodmethod. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255.

46. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1993).Improved methods for building protein models in electron den-sity maps and the location of errors in these models. Acta Crys-tallogr. D Biol. Crystallogr. 49, 148–157.

47. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce bothdetailed and schematic plots of protein structures. J. Appl. Crys-tallogr. 24, 946–950.

48. POV-Team. (1997). POV-Ray: persistence of vision ray-tracer.(http://www.povray.org).

Accession Numbers

Coordinates and structure factors for each structure have beendeposited in the Protein Data Bank (pine STS [1U0U], 18xCHS mu-tant [1U0V], and 18xCHS/resveratrol complex [1U0W]).

chemistry & biology, vol. 11, 1179–1194, september, 2004, 2004 … · 2016-12-01 ·...

Documents